BACKGROUND

Notewell.ai is a HIPA-compliant note-taking tool designed to help social workers document client sessions through voice and AI.

While the product already included an AI chat on desktop, it wasn’t designed to support real workflows: there were no persistent threads, no memory, and no way to tie a conversation to a specific client or meeting. Users had to reload context manually every time, and any chat disappeared as soon as they navigated away.

As we worked to bring Notewell to mobile, we saw a chance to do more than just replicate the feature — we wanted to redesign it from the ground up to feel contextual, connected, and actually useful in the moments social workers needed it most.

discovery

I collaborated with the team to pinpoint the limitations of the desktop AI chat. To guide our direction, we interviewed a few of Notewell’s clients — mostly social workers and program leads — to understand how they used the AI chat, what was missing, and where it could help.

The takeaway was clear: the AI felt disconnected from their actual work. Users wanted to tie questions to specific meetings or clients, revisit past conversations, and rely on the AI to understand context without constant re-explaining.

From there, we explored how the experience could become more persistent, contextual, and naturally integrated into their daily workflows.

what we started with

Before redesigning the AI chat experience, we took a closer look at how it worked on desktop. While the feature was functional, it lacked the structure and context users needed.

From a meeting note, clicking into AI chat opened a separate tab, but context had to be reloaded manually. Threads weren’t saved, client info wasn’t recognized, and chats vanished once the page closed. There was no memory or awareness of the user's workflow.

‍This became the baseline we set out to improve.

RETHINKING THE CHAT MODEL & ENTRY POINTS

At first, AI chat worked the same everywhere, no matter where it was accessed from — always starting a new, temporary thread with no memory or link to a client or meeting. But user interviews made it clear: context was key to making AI useful.

We introduced a new chat model with three entry points, each tailored to a specific need:

• General: Accessed from the main nav. Always opens a new thread that automatically saves. Designed for quick, one-off questions.
• Client-based: Accessed from within a client profile. Threads save and are tied to the context of that specific client. Used for longitudinal questions (e.g. “What themes have come up with this client recently?”).
• Meeting-based: Accessed from a specific meeting note. Threads save and are anchored to that meeting context, ideal for quick summaries or follow-up planning.

INTEGRATING THE AI CHAT INTO THE WORKFLOW

The initial mobile AI chat shipped as part of our broader mobile effort, but it mirrored the desktop version’s flaws. Every request started from scratch, threads vanished on exit, and users had to reload context like meeting notes or clients manually.

Social workers found switching to a separate screen disruptive, so we redesigned the chat as a dedicated tab within client profiles and meeting notes. Anchored to its context, it let users review notes, summarize meetings, and ask follow-ups without losing their place — keeping AI seamlessly in their workflow.

MAKING GENERAL AI CHAT CONTEXT-AWARE

We learned through user interviews that social workers expected the AI to “know” which client or meeting they were referencing, even in general chat. Without that context, responses often felt vague or irrelevant, and adding the necessary details each time was tedious.

Contextual awareness is a core selling point for Notewell, so I designed a lightweight flow that allowed users to attach a client or meeting note before submitting a question.

The initial desktop modal-based approach felt clunky on mobile, so I proposed a bottom sheet — a more native pattern that enabled smoother, in-flow context selection. It also allowed us to scale by supporting both clients and meeting notes as context.

The solution not only improved general chat, it became the standard for adding context across the app.

DESIGNING FOR AI TRUST & CLARITY

Most social workers we spoke to had tried AI tools like ChatGPT but rarely used them for work — citing privacy concerns, vague responses, and the need to constantly re-explain context.

To build trust, we grounded the experience in clarity and transparency. A welcoming message and HIPAA-compliant badge established credibility. Prompt suggestions encouraged exploration, while contextual labels and in-line explanations made it clear how and why the AI was responding — turning an unfamiliar tool into something transparent, helpful and safe to use.

CHATS HISTORY HUB

We initially faced internal hesitation about including a centralized AI chat history. The founder felt the existing entry points — client profiles and meeting notes — were enough, and viewed general chat as a space for quick, disposable questions.

But early feedback made it clear: users wanted to revisit all conversations, including general ones.

Rather than just segmenting history by chat type — which felt clunky — we reframed the challenge: How might we surface past chats effortlessly, regardless of where they started? This led to a unified, lightweight system that felt familiar, flexible, and easy to navigate.

Context-Driven, Trusted AI Chat

The final mobile experience turns AI chat into a reliable, context-rich assistant. Anchored directly within client profiles and meeting notes, conversations persist across sessions and are always tied to the right casework. A unified history surfaces all past chats by recency, making it effortless to pick up where you left off. The result is a tool that feels less like a separate feature and more like an integral part of their daily workflow.